Abstract: Using the flight parameters data recorded in flight training process can identify the flight action type operated by a pilot. The subseries of flight parameters data have characters of temporal relation and correlation. And the parameters have different dimension and high volume data. In view of the reasons, this method makes use of principal component analysis to extract the statistical feature of parameters’ relevance, and makes use of Euclidean distance discriminate analysis to process rude identification of the flight action type. Then, the multivariate time series data of flight parameters which have high relevance are extracted, and their sequence change features are expressed by a section method. Finally, the method of dynamic time warping distance is used to process the feature-fitting classification. Through the meticulous classification, the flight action type is identified. The test proves that the method can improve the correct recognition rate and the efficiency of the identification of flight action type.

Key words: multivariate time series, dynamic time warping, Principal Component Analysis（PCA）, pattern matching

摘要： 利用飞行训练过程中记录的飞行参数数据，可以识别飞行员操作的飞行动作类型。飞行参数数据的各参数子序列具有时序性、相关性、各参数量纲不同、数据量大的特点，所以该方法考虑利用主成分分析提取其参数相关度统计特征，使用欧氏距离判别分析进行粗分类；然后提取相关度较大参数数据的时序趋势变化特征，利用动态时间弯曲距离匹配细分类，进一步识别动作区段归属的飞行动作类型。该方法能够在保证识别率的前提下，提高飞行动作识别的效率。

关键词: 多元时间序列, 动态时间弯曲, 主成分分析, 模式匹配